Exp Brain Res

DOI 10.1007/s00221-010-2496-8

RESEARCH ARTICLE

DTI-MR tractography of white matter damage in stroke patients with neglect

M. Urbanski · M. Thiebaut de Schotten · S. Rodrigo · C. Oppenheim · E. Touzé · J.-F. Méder · K. Moreau · C. Loeper-Jeny · B. Dubois · P. Bartolomeo

Received: 6 July 2010 / Accepted: 9 November 2010© Springer-Verlag 2010

Abstract Left visual neglect is a dramatic neurologicalcondition that impairs awareness of left-sided events.Neglect has been classically reported after strokes in theterritory of the right middle cerebral artery. However, theprecise lesional correlates of neglect within this territoryremain discussed. Recent evidence strongly suggests animplication of dysfunction of large-scale perisylvian net-works in chronic neglect, but the quantitative relationshipsbetween neglect signs and damage to white matter (WM)tracts have never been explored. In this prospective study,we used diVusion tensor imaging (DTI) tractography intwelve patients with a vascular stroke in the right hemi-sphere. Six of these patients showed signs of neglect. Non-parametric voxel-based comparisons between neglect andcontrols on fractional anisotropy maps revealed clusters inthe perisylvian WM and in the external capsule. IndividualDTI tractography identiWed speciWc disconnections of the

fronto-parietal and fronto-occipital pathways in the neglectgroup. Voxel-based correlation statistics highlighted corre-lations between patients’ performance on two visual searchtasks and damage to WM clusters. These clusters werelocated in the anterior limb of the internal capsule and inthe WM underlying the inferior frontal gyrus, along the tra-jectory of the anterior segment of the arcuate fasciculus(asAF). These results indicate that chronic visual neglectcan result from, and correlate with, damage to fronto-parie-tal connections in the right hemisphere, within large-scalecortical networks important for orienting of spatial atten-tion, arousal and spatial working memory.

Keywords Attention · DiVusion tensor imaging tractography · Hemispatial neglect · Stroke

Introduction

Vascular strokes in the right hemisphere often result insigns of neglect for events occurring on the left side of

Electronic supplementary material The online version of this article (doi:10.1007/s00221-010-2496-8) contains supplementary material, which is available to authorized users.

M. Urbanski (&) · B. Dubois · P. Bartolomeo (&)INSERM-UPMC UMR S 975, G.H. Pitié-Salpêtrière, 47 boulevard de l’Hôpital, 75013 Paris, Francee-mail: marika.urbanski@gmail.com

P. Bartolomeoe-mail: paolo.bartolomeo@upmc.fr

M. Thiebaut de SchottenNatbrainlab, Department of Forensic and Neurodevelopmental Sciences, Institute of Psychiatry, King’s College London, London, UK

S. Rodrigo · C. Oppenheim · J.-F. MéderDepartment of Neuroradiology, Hôpital Sainte-Anne, Université Paris Descartes, Paris, France

E. TouzéDepartment of Neurology, Hôpital Sainte-Anne, Université Paris Descartes, Paris, France

M. Urbanski · K. Moreau · C. Loeper-JenyDepartment of Functional Rehabilitation, Hôpital National de Saint-Maurice, Saint-Maurice, France

B. Dubois · P. BartolomeoDepartment of Neurology, AP-HP, IFR 70, Hôpital de la Salpêtrière, Paris, France

P. BartolomeoDepartment of Psychology, Catholic University, Milan, Italy

123

Exp Brain Res

patients’ space. Neglect patients seem to live in a halvedworld: they do not eat from the left part of their dish, orbump their body into obstacles situated on their left. Whenreproducing a linear drawing, they fail to copy the left partof the whole scene or of objects therein. On the other hand,patients’ gaze tends to be captured by right-sided, ipsile-sional objects, as if they exerted a sort of “magnetic” attrac-tion (Gainotti et al. 1991). Besides its obvious interest forthe cognitive neuroscience of visual attention and spatialprocessing, a better understanding of visual neglect isimportant on clinical grounds. In particular, post-strokefunctional recovery is often poor in these patients (Malho-tra et al. 2006) despite available rehabilitation procedures(Pisella et al. 2006) and promising possibilities for pharma-cological treatments (Coulthard et al. 2008).

In the great majority of cases, lesions associated withneglect are localised in the territory of the middle cerebralartery (Husain and Nachev 2007; Mort et al. 2003). Toidentify the precise lesional correlate of neglect, studieshave most often used the lesion overlapping method: thelesions of patients with neglect are superimposed in a refer-ential space, lesions of brain-damaged patients withoutsigns of neglect are subtracted out and the locus of maxi-mum overlapping is considered as the critical region whosedamage produces neglect. These studies reported hotspotsin several distinct cortical loci: the angular gyrus in theinferior parietal lobule (Mort et al. 2003), portions of theangular and supramarginal gyri at the junction with thetemporal lobe (Vallar 2001), more rostral portions of thesuperior temporal gyrus (Karnath et al. 2004), or the infe-rior frontal gyrus (Husain and Kennard 1996). Studiesbased on voxel-based lesion-symptom mapping (VLSM;see Bates et al. 2003) have also shown correlations betweenthe involvement of the right inferior frontal gyrus andneglect signs (Committeri et al. 2007; Verdon et al. 2010).Thus, in most cases, lesions appear to cluster around a largeperisylvian network in the right hemisphere (Mesulam1981; Heilman et al. 1983; Bartolomeo 2006, 2007).

Signs of left neglect are likely to result from the interac-tion of spatial and non-spatial deWcits (Bartolomeo 2007;Husain and Nachev 2007). Among these components, deW-cits of spatial attention have often been stressed (Posneret al. 1984; Losier and Klein 2001; Bartolomeo and Chok-ron 2002). fMRI evidence in normal participants indicatethat orienting of spatial attention depends on the coordi-nated activity of fronto-parietal networks (Nobre 2001;Corbetta and Shulman 2002). In left neglect, left-hemi-sphere unimpaired networks show abnormal activity (Corb-etta et al. 2005) and TMS-induced suppression of thesenetworks ameliorates neglect (Koch et al. 2008). Neglectpatients also often show non-lateralized deWcits, such as animpairment of spatial working memory and reduced arousal(Husain and Rorden 2003).

In keeping with the multifarious nature of their symp-toms, patients with neglect often have relatively largelesions, which are likely to disrupt several functionalmodules. If so, however, the lesion overlapping method,with its emphasis on focal hotspots, might not be conduciveto accurate anatomo-clinical correlations. Moreover, thevoxel-based statistics used by the lesion overlappingmethod relies on the “topological” assumption (Catani andFfytche 2005) that the voxels of maximum overlap corre-spond to the crucial cortical correlate of the neurologicaldeWcit. While this may well be the case, it is also possible,according to a more “hodological” perspective, that lesionsplaced on diVerent locations along the trajectory of a whitematter pathway impair the integrated functioning of thecortical network connected by that pathway (Catani andMesulam 2008). In this case, the lesion overlap methodwould be clearly inadequate to identify the brain network atissue. Along these lines, recent studies employed methodsdiVerent from lesion overlapping to investigate the neuralbases of neglect. Combining diVusion tensor imaging (DTI)tractography (Basser et al. 1994) with direct electrical stim-ulation of the brain (DuVau et al. 1999), Thiebaut de Schot-ten et al. (2005) showed that the temporary inactivation ofthe likely human homologue of the second branch of thesuperior longitudinal fasciculus (SLF II), a fronto-parietalwhite matter pathway (Schmahmann and Pandya 2006),can provoke transitory signs of left neglect. This evidenceconWrmed and speciWed the Wndings of Leibovitch et al.(1998) and Doricchi and Tomaiuolo (2003), who reported amaximum lesion overlap on the SLF in stroke patients withneglect. A maximum lesion overlap on white matter wasalso reported in the relatively rare cases of neglect afterlesions in the territory of the right posterior cerebral artery(Mort et al. 2003; Park et al. 2006). The overlap location wascompatible with the trajectory of the inferior longitudinal fas-ciculus (ILF; Bird et al. 2006). Using DTI tractography,Urbanski et al. (2008) performed in vivo reconstruction ofthe SLF, the ILF and a further long-range pathway running inthe depth of the temporal lobe, the inferior fronto-occipitalfasciculus (IFOF), in four patients with right brain lesions.DTI evidence of IFOF disconnection was present only in thetwo patients showing signs of left neglect.

DiVusion tensor imaging tractography has recently beenused to obtain a detailed anatomical description of thewhite matter pathways connecting perisylvian brain areasin the left human hemisphere (Catani et al. 2005). There isa well-established direct pathway, the arcuate fasciculus(AF), connecting the Wernicke territory (including the pos-terior part of both the superior temporal gyrus, STg andmiddle temporal gyrus, MTg) with the Broca territory(Brodmann areas 44 and 45, and part of the middle frontalgyrus and inferior precentral gyrus). This long segment pre-sents an asymmetry favouring the left hemisphere, perhaps

123

Exp Brain Res

related to its role in language processes (Rodrigo et al.2007). In addition, there is an indirect pathway consistingof two segments, a posterior segment (psAF) connectingWernicke territory with the inferior parietal lobe (IPL),and a fronto-parietal or anterior segment (asAF) linkingIPL with posterior Broca territory (compared to the pro-jections of the AF), which likely corresponds to thehuman homologue of the third branch of the superior lon-gitudinal fasciculus (SLF III; Schmahmann and Pandya2006). However, DTI tractography showed that in theright hemisphere, the long segment appears to be presentonly in 40% of the normal population (Catani et al. 2007).Therefore, 60% of people seem to have only the asAF(connecting IPL and posterior Broca territory) and thepsAF (connecting IPL and STg/MTg) in the right hemi-sphere (Catani et al. 2007). Interestingly, results of Cataniet al.’s study (2007) indicated an inter-hemispheric diVer-ence in the structural organisation of the asAF, withhigher fractional anisotropy (FA) values in the right hemi-sphere than in the left hemisphere, consistent with thepossibility of a right-hemisphere ventral attentional net-work (VAN; Corbetta and Shulman 2002).

Indeed, besides their crucial role for language in the lefthemisphere, these networks are also important for attentionalprocesses. The right-hemisphere homologue of the networkconnected by asAF/SLF III and psAF is active when subjectsreorient their attention from an expected location to an unex-pected one (Corbetta and Shulman 2002). More dorsal areasin the posterior parietal cortex (PPC) and in the lateral pre-frontal cortex (PFC), connected by more dorsal branches ofthe SLF (Schmahmann and Pandya 2006; Thiebaut de Schot-ten et al. 2008), are active during spatial orienting tasks(Nobre 2001; Corbetta and Shulman 2002). In particular,PPC activity is modulated by area 46 in the PFC (McIntoshet al. 1994; Büchel and Friston 1997). The human PPC (area7 extending in the intraparietal sulcus) modulate responses inV5/MT area (analogue of area MT in the monkey) duringattentional tasks (Friston and Büchel 2000). This suggeststhat the psAF has a role in synchronising activity in the dor-sal visual pathway. By connecting parietal and temporal cor-tical modules, the posterior segment may also permit thetime-locked integration of spatial and perceptual informationthat is necessary for attentional selection and conscious pro-cessing of visual objects (see Robertson 2003).

Concerning the white matter pathways ventral to theperisylvian regions, whose disconnection has also beenimplicated in neglect, the ILF is a ventral associative bun-dle with long and short Wbres connecting the occipital andtemporal lobes. The long Wbres, which run medially to theshort Wbres, connect visual areas to the amygdala and thehippocampus (Catani and Thiebaut de Schotten 2008). TheIFOF connects the ventro-lateral pre-frontal cortex and medialorbitofrontal cortex to the occipital lobe (Catani et al. 2002).

The optic radiations (OR), linking the lateral geniculatenucleus to the primary visual cortex, constitute a furtherwhite matter tract of interest in neglect, because their lesioncan determine visual Weld defects. Homonymous hemiano-pia and visual neglect double dissociate in diVerentpatients. However, left hemianopia can interact withneglect in determining patients’ performance, for exampleby dramatically increasing rightwards shifts of the subjec-tive midpoint in line bisection (Doricchi and Angelelli1999).

In view of these notions, it is important to explore thelesional correlates of neglect in stroke patients, by takinginto proper account possible hodological determinants ofneglect. To this aim, we obtained anatomical and DTIimages in a group of 12 patients with right-hemisphere vas-cular lesions, six of whom showed signs of left neglect, andin 12 age-matched controls without neurological history.After traditional lesion overlapping, for each individualparticipant we performed a virtual in vivo dissection of theperisylvian networks, ILF, IFOF and OR. We subsequentlycombined tract dissection with a voxel-based approach(Ashburner and Friston 2000; Good 2001). Finally, we cal-culated the correlations between patients’ performance onneuropsychological tests for neglect and DTI-based mea-sures of structural integrity of the explored tracts.

Method

Participants

Twelve controls (6 women and 6 men; mean age59.70 years, SD 7.52) without neurological history andtwelve patients (4 women and 8 men) with a vascular strokein the right hemisphere participated in this study. All partic-ipants were right handed and gave written informed con-sent. The ethics committee of the Hôtel-Dieu Hospital inParis approved the protocol. Patients were consecutivelyincluded on the basis of inclusion and exclusion criteria.Inclusion criteria were as follows: age 18–80 years, pres-ence of a unique stroke in the right hemisphere, absence ofother white matter pathology (e.g. leukoaraiosis), timeinterval of at least 3 weeks between stroke onset and MRI.The exclusion criteria were the following: previous stroke,impaired comprehension, psychiatric disorders, altered vig-ilance, contraindications to MRI. Neglect was assessedusing a paper-and-pencil neglect battery (Bartolomeo andChokron 1999). Six patients presented signs of left visualneglect (N+) (63.75 years, SD 9.20), six other patients didnot show any sign of neglect (N¡) (58.77 years, SD 8.67).In the N+ group, the onset of stroke to time of neglect eval-uation ranged from 9 to 2,187 days; in the N¡ group, theonset ranged from 5 to 571 days. Demographical, clinical

123

Exp Brain Res

and neuropsychological data are reported in Table 1. Due toMRI inclusion criteria, time range between stroke onset andMRI was diVerent in N+ and N¡ groups (20–2,187 days inthe N+ group and 36–577 days in the N¡ group). Therewas no clinical evidence of neglect in the acute phase forN¡ patients. As it is often the case in these studies, neglectpatients appeared to have larger lesion volume (mean,88.38 cm3; SD, 71.55) than non-neglect patients (mean,16.52 cm3; SD, 18.36). However, the diVerence was not

statistically reliable (Wilcoxon–Mann–Whitney test,W = 93.5, P = 0.23), probably as a consequence of the largevariability in both groups.

Neuropsychological evaluation

The neglect battery included a line bisection test consistingin eight lines horizontally disposed in a vertical A4 sheet ina Wxed random order (three 60 mm samples, three 100 mm

Table 1 Clinical and demographical data

a Pathological scores. For the line bisection test, cut-oV at +11% deviation (Bartolomeo et al. 1994); for the bells cancellation, left–right diVerence>2 (Rousseaux et al. 2001); for the line cancellation, number of left omissions >1 (Albert 1973); for the letter cancellation, left–right diVerence >2;for the landscape drawing, score <6; for the overlapping Wgures, left omission >1 (Rousseaux et al. 2001). Visual Welds: N normal, LHH left hom-onymous hemianopia, LE left extinctionb Leftward bias due to the compensation of LHH learned during rehabilitation proceduresc The 4 leftmost lines on the test sheet were omitted (see “Methods” for the corrected score of the deviation). STg superior temporal gyrus, IPLinferior parietal lobule, pMTOg posterior part of the middle temporo-occipital gyrus, pI posterior insula, TP temporal pole, IFg inferior frontalgyrus, Fus fusiform gyrus, H hippocampus, MTg middle temporal gyrus, ITg inferior temporal gyrus, F frontal, SPL superior parietal lobule, pITgposterior part of the inferior temporal gyrus, T temporal, P parietal, STS superior temporal sulcus, BG basal ganglia, CR corona radiata, O occipital,I insula, WM white matter

Lesion site Visual Weld

Gender/age/education (years of schooling)

Onset of illness (days)

Line cancellation left/right hits (max 30/30)

N – 1 pI, STG, IPL, pMTOG N F/45/14 9 30/30

N – 2 pI, TP, STG, MTG, ITG N M/60/14 5 30/30

N – 3 F paraventricular, centrum semiovale N M/54/10 36 30/30

N – 4 Cuneus LHH M/69/12 571 30/30

N – 5 F paraventricular, pPutamen N M/58/12 177 30/30

N – 6 Infarct in the WM underlying the pars opercularis of the IFg N M/66/10 321 30/30

N + 1 T, O, rolandic, thalamo-capsular LHH M/64/10 2,187 17/30a

N + 2 Fus, H, I, TP, T, IFg, lenticular, thalamo-capsular, rolandic operculum, Precentral, Postcentral, SMg

LHH M/66/8 65 0/16a

N + 3 T, P, frontal operculum, STS, IPL, MOG LE F/54/8 18 14/15a

N + 4 IPL, SPL, precuneus, cuneus, MTOG, pITG LHH F/80/17 729 30/30

N + 5 I, lenticular, F paraventricular N M/61/17 120 30/30

N + 6 Subinsular and temporal stem WM, BG, CR, IPL LE F/59/10 9 29/30

Bells cancellation left/right hits (max 15/15)

Letter cancellation left/right hits (max 30/30)

Line bisection (% deviation)

Overlapping Wgures left/right hits (max 10/10)

Landscape drawing (max 6)

N ¡ 1 15/15 29/30 ¡3.10 10/10 6

N ¡ 2 15/15 28/29 4.80 10/10 6

N ¡ 3 12/13 28/29 5.85 10/10 6

N ¡ 4 15/15 29/29 ¡17.10a,b 10/10 6

N – 5 15/15 30/29 8.00 10/10 6

N – 6 15/15 30/30 0.24 10/10 6

N + 1 0/15a 1/22a 20.2a 9/10a 4.5a

N + 2 0/5a 0/22a 86.9a,c 2/5a 1a

N + 3 3/12a 15/25a 11.00 9/10a 4.5a

N + 4 1/15a 9/28a 1.00 9/10a 3.5a

N + 5 7/13a 29/30 14.20a 10/10 5a

N + 6 0/6a 0/13a 15.70a 6/10a 4.5a

123

Exp Brain Res

samples and two 180 mm samples; Bartolomeo et al. 1994);three cancellation tests in which patients were asked to can-cel stimuli of various kind: (1) lines (Albert 1973), (2) Asamong other letters (Mesulam 1985), (3) silhouettes ofbells among other objects (Gauthier et al. 1989); an over-lapping Wgures task in which patients where requested toidentify Wve patterns of overlapping linear drawings ofcommon objects (one central and a pair of objects over eachof its sides); a copy of a linear drawing representing a cen-tral house and four trees (a pair of trees over each of itsside) presented on a horizontal A4 sheet. Visual Welds andvisual extinction were assessed using the confrontationtask, which was administered following a previouslydescribed procedure (Bartolomeo and Chokron 1999).

Patients were considered to show left extinction whenthey failed to report at least one left visual stimulus occur-ring simultaneously with a right one; they were consideredto show left hemianopia when they failed to report all leftvisual stimuli even on single hemiWeld stimulation. Diagno-sis of neglect was based on pathological performance on atleast 3 tests of the neglect battery. Patients were assigned tothe non-neglect group when their performance was patho-logical on no more than one test of the neglect battery.

For the line bisection test, the cumulated percentage ofdeviation from the true centre for all the 8 lines was calcu-lated. Rightward deviation assumed a positive sign (max+100), whereas leftward deviations carried a negative sign(max ¡100). Patient N + 2 (see Table 1) completely omit-ted to bisect the 4 leftmost lines in the test sheet. For theseomitted lines, the deviation score was calculated as if thepatient had put the bisection mark at the right endpoint ofthe line. For the landscape drawing, each completely copiedtree was scored 1 point and the complete house 2 points.Items showing evidence of object-based neglect (i.e. onlythe right part of an item correctly drawn) received a scoreof 0.5 point (see Table 1).

For the correlation analyses between integrity of WMand performance to neuropsychological tests, we computeda laterality score (Bartolomeo and Chokron 1999) for thecancellation tests and for the overlapping Wgures test, whichwas entered as a regressor. This score is deWned as(x1 ¡ x2)/(x1 + x2). Values for x1 were given by the numberof items cancelled (or reported for the overlapping Wgurestask) on the right half of the page; values for x2 corre-sponded to the number of left-sided cancelled items (orreported for the overlapping Wgures task). One advantage ofthis score is that it provides a quantitative estimate of spa-tial bias that is independent of the overall level of perfor-mance (e.g. of the total number of cancelled/reportedtargets). Its possible range is from ¡1 (all the items can-celled/reported on the left side, none on the right side) to +1(the opposite situation). A correction was needed for can-cellation tasks performed by patients with severe neglect,

who cancelled only the rightmost items, without crossingthe midline. In order not to underestimate their neglect, thelaterality score obtained by these patients was augmentedby the proportion of the number of neglected items on theright side (max +1.93, corresponding to a single bell can-celled on the right). For the line bisection test, the cumu-lated percentage of deviation for all the 8 lines was enteredas a regressor (for patient N + 2, the percentage of devia-tion was corrected as shown above). For the landscapedrawing, the score/6 was entered as a regressor.

Magnetic resonance imaging acquisition

An echo-planar imaging at 1.5T (General Electric) with astandard head coil for signal reception was used for all theacquisitions. High-resolution 3-D anatomical SPGR imageswere Wrst acquired for each participant (114 axial contigu-ous images, 1.2 mm thick with a FOV of 28 cm).

DTI axial volumes were obtained using a repetition timeof 6,575 ms with an echo time of 74.3 ms and a Xip angle of90°. DiVusion weighting images were performed along 36independent directions, with a b-value of 700 s/mm2. Weused a slice thickness of 4 mm with no gap. The resolutionof this acquisition sequence was 1.09 £ 1.09, the matrixsize was 256 £ 256 with a FOV of 28 cm. The overallacquisition time took was 620 s.

Lesion analysis

An expert blind to the results of the neglect battery drew alllesions manually on slices of the T1 template from MRIcrosoftware (http://www.mricro.com). Lesions for the N+ andN¡ groups were overlapped separately on the Colin27(Holmes et al. 1998) registered in a stereotaxic space (Mon-treal Neurological Institute, MNI, http://www.mni.mcgill.ca/).

Tract-based analysis

DiVusion toolkit (http://www.trackvis.org/) computed thediVusion tensors for each subject and performed an interpo-lated streamline tractography for all voxels with an FAabove an arbitrary threshold of 0.2. This threshold was cho-sen following two studies, which tested diVerent thresholdswith tractography of the cortico-spinal tract in patients withstroke (Kunimatsu et al. 2004) and the uncinate fasciculusin Alzheimer’s disease (Taoka et al. 2009). An anglethreshold of 45° was chosen in order to reduce artefactualreconstructions. We used a standard two-region of interestapproach to isolate streamlines of the inferior longitudinalfasciculus (ILF), the inferior fronto-occipital fasciculus(IFOF), the posterior segment (psAF) and the anterior seg-ment (asAF) of the arcuate fasciculus for the left and theright hemisphere (Catani et al. 2002, 2005; Catani and

123

Exp Brain Res

Thiebaut de Schotten 2008). For all fasciculi except theoptic radiations (OR), we placed spherical ROIs in eachparticipant’s FA map in the native space. For the OR, ROIswere manually drawn. For the psAF, the Wrst ROI wasplaced caudally in the white matter underlying the Wer-nicke’s territory (including the posterior part of both theSTg and MTg) and the second ROI was placed in the whitematter of the angular gyrus (Geschwind’s territory; Cataniet al. 2005, 2007). For the asAF/SLF III, the Wrst ROI wasplaced in the white matter underlying Broca’s territory(Brodmann areas 44 and 45, and part of the middle frontalgyrus and inferior precentral gyrus) and the second ROIwas placed caudally including the white matter underlyingthe Geschwind’s territory (Catani et al. 2005, 2007). Forthe ILF and the IFOF, the Wrst ROI was placed in the occip-ital white matter: for the ILF, the second ROI was placed inthe white matter underlying the rostral temporal regions(Catani et al. 2003), whereas for the IFOF, the second ROIwas placed rostrally in the white matter of the anterior Xoorof the external capsule (Catani et al. 2002). Concerning theoptic radiations (OR), a Wrst coronal ROI was drawn in theposterior occipital lobe and the second ROI at the apex ofthe Meyer’s loop (Ciccarelli et al. 2003; Thiebaut de Schot-ten et al. 2010). An example of the ROIs and the resultingtractography for each fasciculus in a representative subjectfrom each group is given in the Supplementary Fig. 1.

Sizes of the ROIs were variable between subjects but didnot diVer signiWcantly between groups of participants,whatever the tract of interest and the hemisphere (see Sup-plementary Fig. 2).

We extracted the number of streamlines for each subjectand each reconstructed tract and calculated its 95% inferen-tial conWdence intervals (ICIs; Tryon 2001; Tryon and Lewis2008) for the mean of each condition. The use of ICIsaddresses some of the problems of traditional null hypothesistesting (Tryon 2001). The ICI method permits to inferstatistical diVerence as well as equivalence by providing anintuitive graphic method (Tryon and Lewis 2008). Non-overlapping ICIs indicate statistical diVerence (� = 0.05),whereas ICI overlap denotes statistical equivalence. Statisti-cal indeterminacy occurs when both tests are failed.

Voxel-based analysis

Brainvisa 3.0.1 (http://www.brainvisa.info) created an FAmap for each DTI. These maps were registered to the MNIspace and smoothed with a full half width maximum(FWHM) of 11 mm following standard options provided inSPM5 (http://www.fil.ion.ucl.ac.uk/spm/).

Voxel-based statistics were performed by using theStatistical non Parametric Mapping toolbox (SnPM5b,http://www.sph.umich.edu/ni-stat/SnPM). SnPM5b computesvoxel-by-voxel nonparametric two sample t tests (called

pseudo t-statistic), by using a standard nonparametric multiplecomparisons procedure based on randomisation/permutationtesting (Holmes et al. 1996; Nichols and Holmes 2002). In thecase of group comparisons, subjects cannot be assigned ran-domly, thus leading to make weak distribution assumptions(Nichols and Holmes 2002). The permutation test does notrequire any distributional assumption (Hayasaka and Nichols2003) and is most suitable for designs with low degrees offreedom available for variance estimation (Salmond et al.2002; Winkler et al. 2008). As FA maps have been shown toexhibit non-normality (Jones et al. 2005), the permutationapproach has been recommended for voxelwise analysis ofDTI data (Smith et al. 2007; Goodlett et al. 2009). Wecompared the three groups (controls, N¡, N+) using pseudot-statistics on the FA maps (1,000 permutations; smoothing ofvariance at 8 FWHM). The results were provided at P < 0.05with a correction for multiple comparisons (FWE). Weassessed the co-variation between the integrity of the whitematter and performance on the neuropsychological tests (thescores for the cancellation tests, the overlapping Wgures test,the landscape drawing and the percentage of deviation for theline bisection test) with a simple nonparametric regressionperformed on the FA maps of twelve patients and six controls(three men and three women) who had performed the com-plete paper-and-pencil test battery. The nonparametric regres-sion method implemented in SnPM5b does not assume anyparticular relationship between the variables (see Gosh et al.2007) and permits permutations.

Tract overlap maps

For each tract of each control participant, a binary map wascomputed by assigning each pixel a value of 1 or 0 dependingon whether the pixel was intersected by the tract (Ciccarelliet al. 2003; Thiebaut de Schotten et al. 2008). The 12 binarymaps obtained for each tract were spatially normalised to theFA map computed in the voxel-based analysis. The maps werethen summed in SPM5 (http://www.fil.ion.ucl.ac.uk/spm/) toproduce overlap maps for each tract of interest.

Results

Lesion overlap

Overlay lesion plots of the N+ and the N¡ groups are rep-resented in Fig. 1.

A Wrst maximum lesion overlap (Fig. 1, yellow, 5/6patients) in the N+ group was found in the right external cap-sule (Z = ¡4). The overlap extended to the grey matter and thewhite matter of the insula, the STg and the rolandic operculum(Z = ¡4 to 20). A second overlap (Z = 28) was in the superiorparaventricular white matter (Fig. 1, green, 4/6 patients).

123

Exp Brain Res

In the N¡ group, the lesions overlapped on the rightinsula and the right STg (Z = ¡4) in 3/6 patients, a regionalso found in the N+ group.

Tractography reconstruction

Figure 2 shows the mean track counts in the left and in theright hemisphere with 95% inferential conWdence intervals(ICIs; Tryon 2001) for each subject group and each trackedfasciculus. In the left hemisphere, there was substantial ICIoverlapping for N+, N¡ and controls, indicating statisticalequivalence of the number of streamlines. The only excep-

tion was the asAF/SLF III in the N¡ group, where lessstreamlines were found in the left hemisphere than in theright hemisphere.

In the right hemisphere, ICIs showed substantial overlap-ping for the ILF, which resulted thus comparable in patientsand controls. All neglect patients but N + 5 and N + 6 pre-sented a disconnection of psAF, which resulted in large vari-ability of track counts in the N+ group and consequentstatistical indeterminacy. The neglect group diVered signiW-cantly from both the N¡ and the control groups for threeright hemisphere tracts: OR (patients N + 1, N + 2 and N + 4had complete disconnection, consistent with their left hom-onymous hemianopia), IFOF (all N+ patients presented acomplete disconnection) and asAF/SLF III (patients N + 1,N + 2, N + 3 and N + 5 had a disconnection of this tract).

Table 2 summarises for the neglect patients their perfor-mance to the neglect battery and the tracts disconnected.

Voxel-based analysis

Neglect patients had signiWcantly lower FA level whencompared to controls in the right hemisphere, in clusters(Fig. 3, blue) mostly localised in the white matter underly-ing the pars opercularis of the inferior frontal gyrus, thesupramarginal gyrus, the middle temporal gyrus, the occip-ito-temporal white matter and the internal and the externalcapsules. Table 3 displays the relative coordinates in theMNI space. Comparison of right brain-damaged patientswith and without neglect revealed a cluster of decreased FAin the white matter underlying the pars opercularis of theright IFg (MNI coordinates 34 8 22; Fig. 3, pink), in aregion consistent with the trajectory of the asAF/SLF III arerunning through (see Fig. 4). When compared with con-trols, patients without neglect had decreased FA in the right

Fig. 1 N¡: Overlay lesion plots of the patients with right brain damagewithout spatial neglect (n = 6); N+: Overlay lesion plots of the patientswith spatial neglect (n = 6). The number of overlapping lesions is illus-trated by the colour bar coding increasing frequencies from violet (n = 1)to red (n = 6). MNI z-coordinates of each transverse section are given.The white rectangular window zoomed at z = ¡4 shows a maximal over-lap (5/6 N+) in the external capsule. STg Superior temporal gyrus

Fig. 2 Mean track counts with 95% inferential conWdence intervals inthe left (in grey) and the right hemisphere (in white) in each group ofparticipants (Controls; N¡, N+) for all fasciculi (OR; IFOF; ILF;

psAF and asAF/SLFIII). An example of each fasciculus is shown in a3D reconstruction of a brain

123

Exp Brain Res

external capsule (MNI coordinates 32 2 ¡8; Fig. 3, green).Table 4 displays for each patient (N+ and N¡) the sparingor not of the clusters found in the voxel-based analysis.

Correlation with neglect signs

There was a signiWcant nonparametric correlation (atP < 0.05, FWE-corrected) between reduced white matterintegrity (FA maps) and neglect behaviour on the bells andthe letter cancellation tests, whereas there was no correlationwith the other tests of the neglect battery at this threshold.Figure 5 shows a cluster of statistical signiWcance for the letter

cancellation task (MNI coordinates: 18 6 6) (k = 32; Pseudo-t = 6.20, P = 0.032) and two further clusters for the bells can-cellation, one localised in the white matter of the pars opercu-laris of the right inferior frontal gyrus (MNI coordinates: 2810 28) (k = 93; Pseudo-t = 6.10, P = 0.025) and the other inthe anterior limb of the right internal capsule (MNI coordi-nates: 18 4 8) (k = 39; Pseudo-t = 5.88, P = 0.036). Interest-ingly, this cluster was very close anatomically and statisticallyfrom the local maxima of the signiWcant cluster obtained forthe letter cancellation test (Pseudo-t = 6.12; see Fig. 5).

The correlation coeYcient between the laterality scorefor the letter cancellation and the mean FA value in the

Table 2 Performance on the neglect battery and identiWcation of the tract disconnected for each neglect patient

a Pathological scores

Visual Weld Line bisection Bells cancel Line cancel Letter cancel Overlap Wg Landscape drawing Tract

N + 1 LHH 20.2a 0/15a 17/30a 1/22a 9/10a 4.5a OR, IFOF, psAF, asAF

N + 2 LHH 86.9a 0/5a 0/16a 0/22a 2/5a 1a OR, IFOF, psAF, asAF

N + 3 LE 11.00 3/12a 14/15a 15/25a 9/10a 4.5a IFOF, psAF, asAF

N + 4 LHH 1.00 1/15a 30/30 9/28a 9/10a 3.5a OR, IFOF, psAF

N + 5 N 14.20a 7/13a 30/30 29/30 10/10 5a IFOF, asAF

N + 6 LE 15.70a 0/6a 29/30 0/13a 6/10a 4.5a IFOF

Fig. 3 Regions showing signiWcantly reduced FA, in Blue, [Controls]vs. [N+]; in Pink, [N¡] vs. [N+]; in Green, [Controls] vs. [N¡] (allcomparisons P < 0.05 FWE-corrected). The left part The left part ofthe Wgure corresponds to the sagittal, coronal and axial views of a glassbrain representing the pseudo-t statistic in SnPM5b obtained for eachcomparison between groups of participants. The right part of the Wgure

shows the overlap of the statistic maps onto a ch2 template of mricron(http://www.mricron.com) in the MNI coordinates (X Y Z) correspond-ing to the maximal pseudo-t values for the FA diVerence between [N+]and [N¡] (34 8 22) and between [N¡] and [Controls] (32 2 ¡8). IFg(operc.), pars opercularis of the inferior frontal gyrus. MTg Middletemporal gyrus, SMg supramarginal gyrus

123

Exp Brain Res

anterior limb of the right internal capsule (ALIC) wasR2 = 0.73; the correlation coeYcients between the lateralityscore for the bells cancellation and the mean FA value inthe ALIC was R2 = 0.68 and R2 = 0.73 for the mean FAvalue in the WM underlying the pars opercularis of theright IFg (see graphs at the bottom of Fig. 5).1

Discussion

This prospective study aimed at investigating the anatomi-cal correlates of neglect signs in stroke patients. At vari-ance with previous anatomical studies of neglect, mainlybased on topological assumptions, we also took intoaccount possible hodological factors in a small group ofpatients (Catani and Ffytche 2005). By using lesion over-lapping, a method based on topological assumptions, wefound a maximum overlap not in the cortex but in the whitematter, consistent with many previous studies (Doricchiand Tomaiuolo 2003; Bird et al. 2006; Park et al. 2006;

1 These correlations should be interpreted with caution because theyare calculated across groups with non-overlapping performance (withand without neglect) that may drive an artefactual correlation. How-ever, visual inspection of the graphs in Fig. 5 shows that patients withmore severe micro-structural damage tend to show more severe neglecton cancellation tests.

Table 3 White matter regions with signiWcantly decreased FA in all comparisons (FWE-corrected)

Coordinates k Pseudo-t

x y z

N+ vs. controls

IFg (pars opercularis) WM 30 4 22 4,661 9.17***

SMg WM 34 ¡30 28 8.77***

Occipito-temporal WM 34 ¡40 12 7.68**

Internal capsule 18 0 10 5.66*

External capsule 32 2 ¡6 145 6.71**

Middle temporal gyrus 60 ¡50 2 7 4.90*

Inferior frontal gyrus (pars opercularis) 56 10 26 11 4.90*

N¡ vs. controls

External capsule 32 2 ¡8 2 5.00*

N+ vs. N¡IFg (pars opercularis) WM 34 8 22 20 5.54** P < 0.05; ** P < 0.01;

*** P < 0.001

Fig. 4 Overlay onto a FA-MNI template of the asAF/SLFIII overlapmap (in yellow–red; yellow corresponding to a higher degree of over-lap) and of the cluster of statistical signiWcance (MNI coordinates 34 8

22) obtained in SnPM5b for the comparison of [N+] vs. [N¡] (in blue).Each slide comprising the cluster is presented in the axial (Z = 19–25),the coronal (Y = 5–11) and the sagittal (X = 31–37) planes

123

Exp Brain Res

Committeri et al. 2007; Golay et al. 2008; Verdon et al.2010). In 3 of 6 patients without neglect, there was an over-lap on the right insula and on the right superior temporalgyrus. This suggests that damage to these regions plays nocrucial role in neglect.

Given the additional limitations of the lesion overlappingmethod in identifying networks and disconnections, we alsoemployed DTI, a new technique that reveals the organisationof the white matter and its integrity. This method permits toestablish anatomo-functional correlations based on whitematter pathways. We chose to explore patients in a chronicstage, which prevented us from recruiting a large patientseries.2 However, despite the small number of patients ineach group, voxel-based nonparametric statistics on the FAmaps extracted from DTI demonstrated reduced structuralintegrity in the perisylvian white matter and external andinternal capsules of the right hemisphere in neglect patients(Fig. 3). Damage to the internal capsule was probably related

to persistent hemiplegia in the neglect group, especially forthe upper limb (Behrens et al. 2003), whereas the lower limbpartially recovered in 3/6 patients. Some of the clusters of FAdiVerence revealed by the voxel-based analysis (Table 4) liedremotely to the lesion in the N+ group, presumably as a resultof wallerian degeneration. Thus, voxel-based analysis of FAvalues permits to demonstrate WM disconnections lying farfrom the lesion, at variance with other methods (e.g. VLSM).

The virtual dissection of each individual tract allowed usto further specify the involved tracts within the perisylviannetwork revealed by the voxel-based approach. Individualtractography was performed using the ROIs described inthe methods, on the axial individual FA maps. The asAF/SLF III resulted to be signiWcantly involved in neglectpatients, whereas the participation of the psAF was morevariable, leading to statistical indeterminacy (Fig. 2). Con-cerning the external capsule area, which referred to moreventral networks, the IFOF, but not the ILF, was discon-nected in our sample of neglect patients.

2 In order to avoid acute ischemic MRI artefacts such as cell swellingor cytotoxic oedema (see Sotak 2002), we established a minimum nec-essary time interval of 3 weeks between stroke and DT-MR acquisi-tion; as a consequence, many patients tested behaviourally could not beincluded because they had left the hospital before any DTI sequencecould be acquired. Time interval between stroke onset and MRI waschosen on the basis of several studies indicating an initial increase ofFA at the acute stage due to cell swelling. At the sub-acute and chronicstages, there is a decreasing of FA in the lesion, due to wallerian degen-eration (Thomalla et al. 2005). This decrease could remain signiWcant(compared to the FA in the homologous controlateral lesion) even6 months after the stroke (Sotak 2002). The second raison was theprobability of frequent consecutive oedema at the acute stage, whichcould have disturbed the sensitivity of the DTI sequence to the lesion.

Table 4 Examination of the clusters of FA diVerence obtained in the voxel-based analysis in the non-neglect and the neglect patients

MNI coordinates and localisation are displayed on the Wrst raw. In each cell, the sparing or not of the cluster is mentioned on the T1/T2 MRI ofeach subject

¡ spared, + lesioned, ab abnormal appearing (usually in the vicinity of the lesion)

IFg WM White matter underlying the inferior frontal gyrus, SMg WM white matter underlying the supramarginal gyrus, O-T WM occipito-temporalwhite matter, IC internal capsule, EC external capsule, MTg middle temporal gyrus

(30 4 22)IFg WM

(34 ¡30 28)SMg WM

(34 ¡40 12)O-T WM

(18 0 10)IC

(32 2 ¡6)EC

(60 ¡50 2)MTg

(56 10 26)IFg

(32 2 ¡8)EC

(34 8 22)IFg WM

N ¡ 1 ¡/¡ ¡/¡ ab/+ ¡/¡ ¡/¡ +/+ ¡/ab ¡/¡ ¡/¡N ¡ 2 ¡/¡ ¡/¡ +/+ ¡/¡ ¡/¡ ab/ab ab/¡ ¡/¡ ¡/¡N ¡ 3 ab/ab ab/ab ¡/¡ ¡/¡ ¡/¡ ¡/¡ ¡/¡ ¡/¡ ab/ab

N ¡ 4 ¡/¡ ¡/¡ ¡/¡ ¡/¡ ¡/¡ ¡/¡ ¡/¡ ¡/¡ ¡/¡N – 5 ¡/¡ ¡/¡ ¡/¡ ¡/¡ ¡/¡ ¡/¡ ¡/¡ ¡/¡ ¡/¡N – 6 ¡/¡ ¡/¡ ¡/¡ ¡/¡ ab/ab ¡/¡ ¡/¡ +/ab ¡/¡N + 1 +/ab +/+ +/+ +/+ +/+ +/ab ¡/¡ ab/+ ab/ab

N + 2 ¡/ab +/+ +/+ +/+ +/ab ab/+ +/+ +/ab ab/ab

N + 3 +/+ +/ab +/ab ¡/¡ ab/ab +/+ ab/+ ab/ab +/+

N + 4 ¡/¡ ab/ab +/+ ¡/¡ ¡/¡ ab/ab ¡/¡ ¡/¡ ¡/¡N + 5 +/+ +/+ ab/+ ¡/¡ ab/+ ¡/¡ ¡/¡ +/ab ab/¡N + 6 +/+ ab/ab ¡/¡ ¡/¡ +/+ ¡/¡ ¡/¡ +/+ ab/ab

Fig. 5 Regions showing a signiWcant correlation between the integrityof the white matter and the laterality scores for visual search tasks(P < 0.05 FWE-corrected). The statistic map was overlaid onto a ch2template of mricron in the MNI coordinates corresponding to the max-imal pseudo-t values. IFg (operc.), pars opercularis of the inferior fron-tal gyrus. ALIC Anterior limb of the internal capsule. Regression linesand correlation coeYcient (R2) between the laterality score obtainedfor the letter and the bells cancellation tests and the mean FA at themaximal pseudo-t values are represented in the graphs below eachcluster of signiWcant nonparametric regression; each non-neglect pa-tient is represented by a black star and each neglect patient by a blankdiamonds

�

123

Exp Brain Res

123

Exp Brain Res

The OR were also damaged in neglect patients, consis-tent with the presence of homonymous hemianopia in halfof our sample. Although hemianopia can dissociate fromneglect signs, it can also worsen patients’ performancewhen present (Doricchi and Angelelli 1999).

Disconnection of asAF/SLF III is consistent with accu-mulating evidence on the importance of SLF damage inneglect. This evidence comes from animal studies (GaVanand Hornak 1997; Reep et al. 2004), from lesion overlap instroke patients (Doricchi and Tomaiuolo 2003; Thiebaut deSchotten et al. 2008; Verdon et al. 2010) and from neuro-surgical patients, who showed either transitory deWcitsupon temporary electrical inactivation of the SLF (Thiebautde Schotten et al. 2005), or the occurrence or worsening ofneglect signs after surgical interruption of the SLF (Shino-ura et al. 2009).

The asAF/SLF III connects a ventral attentional network(VAN) network which shows BOLD responses in fMRIwhen participants have to respond to invalidly cued targets(Corbetta and Shulman 2002). The VAN might thus beresponsible for reorienting of attention, whereas a moredorsal fronto-parietal pathway, the dorsal attentional net-work (DAN), probably linked by the human homologue ofSLF II, would orient spatial attention during valid cueing.Corbetta et al. (2008) made the further proposal that theDAN, with possible contribution from other prefrontalregions such as the anterior cingulate and the anteriorinsula, Wltrate the activation of the VAN and gate the sen-sory responses according to their behavioural relevance Inneglect patients, damage to right-hemisphere VAN couldcause a functional imbalance between the left and rightDANs, with a hyperactivity of the left dorsal fronto-parietalnetwork, which would provoke an attentional bias towardsright-sided objects (Kinsbourne 1993) and neglect of left-sided items. Consistent with this hypothesis, suppressiveTMS on the left parieto-motor pathway correlated with animprovement of patients’ performance on cancellation tests(Koch et al. 2008). The IPL and its ventral frontal projec-tions can also be responsible for maintaining attention ongoal or task, which is a top-down process (Singh-Curry andHusain 2009). Impaired sustained attention combined witha deWcit in detecting salient events after right hemisphericstroke may lead to an exacerbation of the spatial bias (Hus-ain and Rorden 2003; Husain and Nachev 2007).

We previously described the involvement of the rightIFOF in two neglect patients with predominantly subcorti-cal lesion (Urbanski et al. 2008). This result was conWrmedin the present sample, because the IFOF was disconnectedin all patients with neglect and normal and symmetrical inall patients without neglect. Although damage to the IFOFmight not be necessary by itself to produce signs of neglect[for example, the IFOF was intact in both the patients withneglect and SLF damage described by Shinoura et al.

(2009)], it might contribute to neglect signs by deprivingvisual cortex of top-down modulation from more anteriorregions, or by decreasing the inXuence of visual input onthe right VLPFC, with consequent deterioration of patients’level of arousal (Doricchi et al. 2008; Urbanski et al. 2008)or sustained attention (see Singh-Curry and Husain 2009).

Several mechanisms of compensation, not mutuallyexclusive, are possible after white matter damage (DuVau2009). These include the recruitment of redundant neuronslocated closely to the lesion; the recruitment of accessorycontralesional pathways thanks to the suppression of callo-sal inhibition; the recruitment of parallel long-distanceassociation pathways, whereby direct and indirect intra-hemispheric pathways might compensate for each other. Asa consequence, lesion volume, especially in the white mat-ter, may be a critical factor of chronic neglect, by prevent-ing these adaptive mechanisms to take place (Bartolomeoet al. 2007).

To obtain more quantitative estimates of the relationshipsbetween lesion sites and signs of neglect, we calculated non-parametric correlations between FA values and patients’performance on tests of the neglect battery. We found sig-niWcant correlations between patients’ scores on two paper-and-pencil tests, which required target/distractor discrimina-tion (the letter and the bells cancellation tasks) and reducedstructural integrity in the white matter in a common regionconsistent with the anterior limb of the internal capsule(Fig. 5). The anterior limb of the internal capsule containsthe anterior thalamic radiations, a bundle of Wbres linkinganterior and dorsomedial thalamic nuclei with the prefrontalcortex and the cingulate gyrus, and fronto-pontine Wbres(Crosby et al. 1962). Moreover, the performance in the bellscancellation correlated with another cluster in the white mat-ter underlying the inferior frontal gyrus pars opercularis(Fig. 5), in a location consistent with the trajectory of theasAF/SLF III (Fig. 4). The right IFg, a cortical terminationof the asAF/SLF III, has been implicated in signs of neglectwhen a target/distractor discrimination is needed (Husainand Kennard 1996), which was indeed the case with thebells cancellation test we employed. Interestingly, the recentstudy from Verdon et al. (2010) using a VLSM analysis on80 patients with a right hemisphere stroke has shown thatthe factor (revealed by a prior factorial analysis on patients’performance to a battery of neglect tests) accounting foromission of targets in the bells cancellation test and in theOta search task correlated with a peak in the right inferiorfrontal gyrus (BA 45) and in other prefrontal areas.

We found no signiWcant correlation in our samplebetween line bisection performance and FA values. Thispattern of results is consistent with the suggestion thatpatients with frontal or deep lesions may show neglect oncancellation task but perform normally on line bisection(Binder et al. 1992).

123

Exp Brain Res

Patient N ¡ 4 in the present series presented a patholog-ical leftward bias in line bisection test, perhaps because ofovercompensation (see Robertson et al. 1994). To test thehypothesis that the performance of this particular patientdetermined some of the observed correlations, we per-formed a further correlation analysis after having excludedpatient N ¡ 4. Voxel-based analysis still showed a diVer-ence in the FA integrity between controls and non-neglectpatients located in the external capsule (P < 0.05, FWE-corrected), suggesting that this result was not driven by thelocation of the stroke in this patient. No correlation wasfound between reduced integrity of the white matter andperformance to line bisection test, suggesting that theabsence of correlation presented with all twelve patientswas not driven by the pathological leftward deviation ofthis non-neglect patient. Moreover, correlation with bothcancellation tests (letter and bells cancellation) was stillpresent at P < 0.05, FWE-corrected, even after the exclu-sion of N ¡ 4, in clusters localised very close to thosedescribed in the analyses performed with all patients.

Contrary to the present hodological approach to neglectanatomy, it has recently been argued that white matter dam-age is relatively unimportant in neglect. Using a probabilis-tic cytoarchitectonic atlas based on histological Wndings in10 adult postmortem brains (Jülich atlas; Bürgel et al.2006), Karnath et al. (2009) reanalysed their previouslystudied 140 right brain–damaged patients (78 with neglectand 62 without neglect) (Karnath et al. 2004), by combin-ing the statistical lesion map obtained from the voxel-basedlesion-behaviour mapping with the probabilistic cytoarchi-tectonic maps of the Jülich atlas. They showed that only36.9% of the damaged voxels were in the white matter.Moreover, if damage to perisylvian pathways was typicallyfound in neglect patients (especially in the superior occip-ito-frontal fasiculus, the SLF and the IFOF), the relation-ship between neglect signs and the involvement of thesefasciculi was however not strongly predictive of neglect.However, as Bürgel et al. (2006) acknowledge, their atlasunderestimate the tracts running rostro-caudally, includingthe fronto-parietal connections. Moreover, at variance toour study, Karnath and co-workers studied patients in theacute stage of their stroke. These issues limit the generalityof the conclusions of the Karnath et al’s reanalysis.

There are also limitations in our study. As mentionedbefore, the relatively large slice thickness may haveincreased partial volume eVect, leading to false positivesresults in the voxel-based analysis and false negatives in thetractography reconstruction. Moreover, Wbre crossing, kiss-ing or fanning are limitations for the tensor model used inDTI to well reconstruct and visualise the trajectory of dor-sal bundles such as the SLF I and SLF II. In areas withischaemic injury, FA decreases after a few days in theperilesional areas (see Thomalla et al. 2005) leading to

possible biased DTI-tractography reconstructions. Wechose to study patients in the chronic stage of their infarctsto avoid this limitation, although adaptive mechanismsmight have aVected neglect deWcits (but note that patientswith acute infarcts can instead suVer from transient ischae-mic penumbra in nearby areas and diaschisis phenomena indistant locations). Future studies based on a larger group ofpatients and new algorithms resolving multiple white mat-ter orientations within voxels will be necessary to conWrmand extent our analysis to supplementary tracts (Dell’acquaet al. 2010) and to understand the relationship betweenMRI signs of damage and severity of neglect.

In conclusion, this study is a Wrst attempt to proposeways to explore causal relationship between network dis-connection and visual neglect. The present results convergewith accumulating previous evidence that signs of chronicleft neglect can result from disruption of the coordinatedactivity of large-scale fronto-parietal networks in the righthemisphere.

Acknowledgments Supported by grants from the AP-HP (interfaceprogramme) and the Université Pierre et Marie Curie, Paris 6 (BonusQualité Recherche) to PB.

References

Albert ML (1973) A simple test of visual neglect. Neurology 23:658–664

Ashburner J, Friston KJ (2000) Voxel-based morphometry—the meth-ods. NeuroImage 11:805–821

Bartolomeo P (2006) A parietofrontal network for spatial aware-ness in the right hemisphere of the human brain. Arch Neurol63:1238–1241

Bartolomeo P (2007) Visual neglect. Curr Opin Neurol 20:381–386Bartolomeo P, Chokron S (1999) Egocentric frame of reference: its

role in spatial bias after right hemisphere lesions. Neuropsycholo-gia 37:881–894

Bartolomeo P, Chokron S (2002) Orienting of attention in left unilate-ral neglect. Neurosci Biobehav Rev 26:217–234

Bartolomeo P, D’Erme P, Gainotti G (1994) The relationship betweenvisuospatial and representational neglect. Neurology 44:1710–1714

Bartolomeo P, Thiebaut de Schotten M, Doricchi F (2007) Left unilat-eral neglect as a disconnection syndrome. Cereb Cortex 17:2479–2490

Basser PJ, Mattiello J, LeBihan D (1994) MR diVusion tensor spectros-copy and imaging. Biophys J 66:259–267

Bates E, Wilson S, Saygin A, Dick F, Sereno M, Knight RT, DronkersN (2003) Voxel-based lesion-symptom mapping. Nat Neurosci6:448–450

Behrens TE, Johansen-Berg H, Woolrich MW, Smith SM, Wheeler-Kingshott CA, Boulby PA, Barker GJ, Sillery EL, Sheehan K,Ciccarelli O, Thompson AJ, Brady JM, Matthews PM (2003)Non-invasive mapping of connections between human thalamusand cortex using diVusion imaging. Nat Neurosci 6:750–757

Binder JR, Marshall R, Lazar R, Benjamin J, Mohr JP (1992) Distinctsyndromes of hemineglect. Arch Neurol 49:1187–1194

Bird C, Malhotra P, Parton A, Coulthard EJ, Rushworth MF, Husain M(2006) Visual neglect after right posterior cerebral artery infarc-tion. J Neurol Neurosurg Psychiatr 77:1008–1012

123

Exp Brain Res

Büchel C, Friston K (1997) Modulation of connectivity in visual path-ways by attention: cortical interactions evaluated with structuralequation modelling and fMRI. Cereb Cortex 7:768–778

Bürgel U, Amunts K, Hoemke L, Mohlberg H, Gilsbach JM, Zilles K(2006) White matter Wber tracts of the human brain: three-dimen-sional mapping at microscopic resolution, topography and inter-subject variability. NeuroImage 29:1092–1105

Catani M, Ffytche D (2005) The rises and falls of disconnection syn-dromes. Brain 128:2224–2239

Catani M, Mesulam M (2008) The arcuate fasciculus and the discon-nection theme in language and aphasia: history and current state.Cortex 44:953–961

Catani M, Thiebaut de Schotten M (2008) A diVusion tensor imagingtractography atlas for virtual in vivo dissections. Cortex 44:1105–1132

Catani M, Howard RJ, Pajevic S, Jones DK (2002) Virtual in vivointeractive dissection of white matter fasciculi in the human brain.NeuroImage 17:77–94

Catani M, Jones DK, Donato R, Ffytche D (2003) Occipito-temporalconnections in the human brain. Brain 126:2093–2107

Catani M, Jones DK, Ffytche D (2005) Perisylvian language networksof the human brain. Ann Neurol 57:8–16

Catani M, Allin MP, Husain M, Pugliese L, Mesulam MM, MurrayRM, Jones DK (2007) Symmetries in human brain language path-ways correlate with verbal recall. Proc Natl Acad Sci USA104:17163–17168

Ciccarelli O, Toosy AT, Parker GJ, Wheeler-Kingshott CA, Barker GJ,Miller DH, Thompson AJ (2003) DiVusion tractography basedgroup mapping of major white-matter pathways in the humanbrain. NeuroImage 19:1545–1555

Committeri G, Pitzalis S, Galati G, Patria F, Pelle G, Sabatini U, Cas-triota-Scanderbeg A, Piccardi L, Guariglia C, Pizzamiglio L(2007) Neural bases of personal and extrapersonal neglect in hu-mans. Brain 130:431–441

Corbetta M, Shulman G (2002) Control of goal-directed and stimulus-driven attention in the brain. Nat Rev Neurosci 3:215–229

Corbetta M, Kincade M, Lewis C, Snyder A, Sapir A (2005) Neural ba-sis and recovery of spatial attention deWcits in spatial neglect. NatNeurosci 8:1603–1610

Corbetta M, Patel G, Shulman G (2008) The reorienting system of thehuman brain: from environment to theory of mind. Neuron58:306–324

Coulthard EJ, Nachev P, Husain M (2008) Control over conXict duringmovement preparation: role of posterior parietal cortex. Neuron58:144–157

Crosby EC, Humphrey T, Lauer EW (1962) Correlative anatomy of thenervous system. Macmillian Co., New York

Dell’Acqua F, Scifo P, Rizzo G, Catani M, Simmons A, Scotti G, FazioF (2010) A modiWed damped Richardson–Lucy algorithm to re-duce isotropic background eVects in spherical deconvolution.NeuroImage 49:1446–1458

Doricchi F, Angelelli P (1999) Misrepresentation of horizontal spacein left unilateral neglect: role of hemianopia. Neurology 52:1845–1852

Doricchi F, Tomaiuolo F (2003) The anatomy of neglect without hem-ianopia: a key role for parietal-frontal disconnection? Neurore-port 14:2239–2243

Doricchi F, Thiebaut de Schotten M, Tomaiuolo F, Bartolomeo P(2008) White matter (dis)connections and gray matter (dys)func-tions in visual neglect: gaining insights into the brain networks ofspatial awareness. Cortex 44:983–995

DuVau H (2009) Does post-lesional subcortical plasticity exist in thehuman brain? Neurosci Res 65:131–135

DuVau H, Capelle L, Sichez J, Faillot T, Abdennour L, Law Koune JD,Dadoun S, Bitar A, Arthuis F, Van EVenterre R, Fohanno D(1999) Intra-operative direct electrical stimulations of the central

nervous system: the Salpêtrière experience with 60 patients. ActaNeurochir 141:1157–1167

Friston KJ, Büchel C (2000) Attentional modulation of eVective con-nectivity from V2 to V5/MT in humans. Proc Natl Acad Sci USA97:7591–7596

GaVan D, Hornak J (1997) Visual neglect in the monkey. Representa-tion and disconnection. Brain 120:1647–1657

Gainotti G, D’Erme P, Bartolomeo P (1991) Early orientation of attentiontoward the half space ipsilateral to the lesion in patients with unilat-eral brain damage. J Neurol Neurosurg Psychiatr 54:1082–1089

Gauthier L, Dehaut F, Joanette Y (1989) The bells test: a quantitativeand qualitative test for visual neglect. Int J Clin Neuropsychol11:49–53

Golay L, Schnider A, Ptak R (2008) Cortical and subcortical anatomyof chronic spatial neglect following vascular damage. BehavBrain Funct 4:43

Good CD (2001) Cerebral asymmetry and the eVects of sex and hand-edness on brain structure: a voxel-based morphometric analysis of465 normal adult human brains. NeuroImage 14:685–700

Goodlett CB, Fletcher P, Gilmore JH, Gerig G (2009) Group analysisof DTI Wber tract statistics with application to neurodevelopment.NeuroImage 45:S133–S142

Gosh S, Rao PS, De G, Majumder PP (2007) A nonparametric regres-sion-based linkage scan of rheumatoid factor-IgM using sib-pairsquared sums and diVerences. BMC Proc 1:S99

Hayasaka S, Nichols T (2003) Validating cluster size inference: ran-dom Weld and permutation methods. NeuroImage 20:2343–2356

Heilman KM, Watson RT, Bower D, Valenstein E (1983) Right hemi-sphere dominance for attention. Rev Neurol (Paris) 139:15–17

Holmes AP, Blair RC, Watson JD, Ford I (1996) Nonparametric anal-ysis of statistic images from functional mapping experiments.J Cereb Blood Flow Metab 16:7–22

Holmes CJ, Hoge R, Collins L, Woods R, Toga AW, Evans AC (1998)Enhancement of MR images using registration for signal averag-ing. J Comput Assist Tomogr 22:324–333

Husain M, Kennard C (1996) Visual neglect associated with frontallobe infarction. J Neurol 243:652–657

Husain M, Nachev P (2007) Space and the parietal cortex. TrendsCogn Sci 11:30–36

Husain M, Rorden C (2003) Non-spatially lateralized mechanisms inhemispatial neglect. Nat Rev Neurosci 4:26–36

Jones DK, Symms MR, Cercignani M, Howard RJ (2005) The eVect ofWlter size on VBM analyses of DT-MRI data. NeuroImage26:546–554

Karnath HO, Fruhmann Berger M, Küker W, Rorden C (2004) Theanatomy of spatial neglect based on voxelwise statistical analysis:a study of 140 patients. Cereb Cortex 14:1164–1172

Karnath HO, Rorden C, Ticini LF (2009) Damage to white matter Wbertracts in acute spatial neglect. Cereb Cortex 19:2331–2337

Kinsbourne M (1993) Orientational bias model of unilateral neglect:evidence from attentional gradients within hemispace. In: Robert-son IH, Marshall JC (eds) Unilateral neglect: clinical and experi-mental studies. Lawrence Erlbaum Associates, Hove, pp 63–86

Koch G, Oliveri M, Cheeran B, Ruge D, Lo Gerfo E, Salerno S, Torri-ero S, Marconi B, Mori F, Driver J, Rothwell JC, Caltagirone C(2008) Hyperexcitability of parietal-motor functional connectionsin the intact left-hemisphere of patients with neglect. Brain131:3147–3155

Kunimatsu A, Aoki S, Masutani Y, Abe O, Hayashi N, Mori H, Ma-sumoto T, Ohtomo K (2004) The optimal trackability threshold offractional anisotropy for diVusion tensor tractography of the cor-ticospinal tract. Magn Reson Med Sci 3:11–17

Leibovitch FS, Black SE, Caldwell CB, Ebert PL, Ehrlich LE, SzalaiJP (1998) Brain-behavior correlations in hemispatial neglect us-ing CT and SPECT: the Sunnybrook stroke study. Neurology50:901–908

123

Exp Brain Res

Losier BJ, Klein RM (2001) A review of the evidence for a disengagedeWcit following parietal lobe damage. Neurosci Biobehav Rev25:1–13

Malhotra P, Parton AD, Greenwood R, Husain M (2006) Noradrener-gic modulation of space exploration in visual neglect. Ann Neurol59:186–190

McIntosh AR, Grady CL, Ungerleider LG, Haxby JV, Rapoport SI,Horwitz B (1994) Network analysis of cortical visual pathwaysmapped with PET. J Neurosci 14:655–666

Mesulam M (1981) A cortical network for directed attention and uni-lateral neglect. Ann Neurol 10:309–325

Mesulam M (1985) Principles of behavioral neurology. F.A. Davis,Philadelphia

Mort DJ, Malhotra P, Mannan SK, Rorden C, Pambakian A, KennardC, Husain M (2003) The anatomy of visual neglect. Brain126:1986–1997

Nichols TE, Holmes AP (2002) Nonparametric permutation tests forfunctional neuroimaging: a primer with examples. Hum BrainMapp 15:1–25

Nobre AC (2001) The attentive homunculus: now you see it, now youdon’t. Neurosci Biobehav Rev 25:477–496

Park KC, Lee BH, Kim EJ, Shin MH, Choi KM, Yoon SS, Kwon SU,Chung CS, Lee KH, Heilman KM, Na DL (2006) DeaVerentation-disconnection neglect induced by posterior cerebral artery infarc-tion. Neurology 66:56–61

Pisella L, Rode G, Farnè A, Tilikete C, Rossetti Y (2006) Prism adap-tation in the rehabilitation of patients with visuo-spatial cognitivedisorders. Curr Opin Neurol 19:534–542

Posner MI, Walker JA, Friedrich FJ, Rafal RD (1984) EVects of parietalinjury on covert orienting of attention. J Neurosci 4:1863–1874

Reep RL, Corwin JV, Cheatwood JL, Van Vleet TM, Heilman KM,Watson RT (2004) A rodent model for investigating the neurobi-ology of controlateral neglect. Cogn Behav Neurol 17:191–194

Robertson LC (2003) Binding, spatial attention and perceptual aware-ness. Nat Rev Neurosci 4:93–102

Robertson IH, Halligan PW, Bergego C, Homberg V, Pizzamiglio L,Weber E, Wilson BA (1994) Right neglect following right hemi-sphere damage? Cortex 30:199–213

Rodrigo S, Naggara O, Oppenheim C, Golestani N, Poupon C, Cointe-pas Y, Mangin JF, Le Bihan D, Meder JF (2007) Human subinsu-lar asymmetry studied by diVusion tensor imaging and Wbertracking. AJNR Am J Neuroradiol 28:1526–1531

Rousseaux M, Beis JM, Pradat-Diehl P, Martin Y, Bartolomeo P, Ber-nati T, Chokron S, Leclercq M, Louis-Dreyfus A, Marchal F, Pe-rennou D, Prairial C, Rode G, Samuel C, SieroV E, Wiart L,Azouvi P (2001) Presenting a battery for assessing spatial neglect.Norms and eVects of age, educational level, sex, hand and lateral-ity. Rev Neurol (Paris) 157:1385–1400

Salmond CH, Ashburner J, Vargha-Khadem F, Connelly A, GadianDG, Friston K (2002) Distributional assumptions in voxel-basedmorphometry. NeuroImage 17:1027–1030

Schmahmann JD, Pandya D (2006) Fiber pathways of the brain.Oxford University Press, New York

Shinoura N, Suzuki Y, Yamada R, Tabei Y, Saito K, Yagi K (2009)Damage to the right superior longitudinal fasciculus in the infe-

rior parietal lobe plays a role in spatial neglect. Neuropsychologia47:2600–2603

Singh-Curry V, Husain M (2009) The functional role of the inferiorparietal lobe in the dorsal and ventral stream dichotomy. Neuro-psychologia 47:1434–1448

Smith SM, Johansen-Berg H, Jenkinson M, Rueckert D, Nichols T,Miller KL, Robson MD, Jones DK, Klein JC, Bartsch AJ, BehrensTE (2007) Acquisition and voxelwise analysis of multisubjectdiVusion data with Tract-Based Spatial Statistics. Nat Protoc2:499–504

Sotak CH (2002) The role of diVusion tensor imaging in the evaluationof ischemic brain injury—a review. NMR Biomed 15:561–569

Taoka T, Morikawa M, Akashi T, Miyasaka T, Nakagawa H, KiuchiK, Kishimoto T, Kichikawa K (2009) Fractional anisotropy—threshold dependence in tract-based diVusion tensor analysis:evaluation of the uncinate fasciculus in Alzheimer disease. AJNRAm J Neuroradiol 30:1700–1703

Thiebaut de Schotten M, Urbanski M, DuVau H, Volle E, Levy R,Dubois B, Bartolomeo P (2005) Direct evidence for a parietal-frontal pathway subserving spatial awareness in humans. Science309:2226–2228

Thiebaut de Schotten M, Kinkingnéhun S, Delmaire C, Lehéricy S,DuVau H, Thivard L, Volle E, Levy R, Dubois B, Bartolomeo P(2008) Visualization of disconnection syndromes in humans. Cor-tex 44:1097–1103

Thiebaut de Schotten M, Ffytche D, Bizzi A, Dell’Acqua F, Allin M,Walshe M, Murray R, Williams SC, Murphy DGM, Catani M(2010) Atlasing location, asymmetry and inter-subject variabilityof white matter tracts in the human brain with MR diVusion trac-tography, NeuroImage. doi:10.1016/j.neuroimage.2010.07.055

Thomalla G, Glauche V, Weiller C, Rother J (2005) Time course ofwallerian degeneration after ischaemic stroke revealed by diVu-sion tensor imaging. J Neurol Neurosurg Psychiatry 76:266–268

Tryon WW (2001) Evaluating statistical diVerence, equivalence, andindeterminacy using inferential conWdence intervals: an inte-grated alternative method of conducting null hypothesis statisticaltests. Psychol Methods 6:371–386

Tryon WW, Lewis C (2008) An inferential conWdence interval methodof establishing statistical equivalence that corrects Tryon’s (2001)reduction factor. Psychol Methods 13:272–277

Urbanski M, Thiebaut de Schotten M, Rodrigo S, Catani M, Oppen-heim C, Touze E, Chokron S, Meder JF, Levy R, Dubois B, Bar-tolomeo P (2008) Brain networks of spatial awareness: evidencefrom diVusion tensor imaging tractography. J Neurol NeurosurgPsychiatry 79:598–601

Vallar G (2001) Extrapersonal visual unilateral spatial neglect and itsneuroanatomy. NeuroImage 14:S52–S58

Verdon V, Schwartz S, Lovblad KO, Hauert CA, Vuilleumier P (2010)Neuroanatomy of hemispatial neglect and its functional compo-nents: a study using voxel-based lesion-symptom mapping. Brain133:880–894

Winkler AM, Nichols T, Glahn DC (2008) On non-normality, non-parametric tests and pooling permutations over space for voxel-based morphometry. Poster presented at the congress humanbrain mapping, Melbourne, Australia

123